Most Regulatory Interactions Are Not in Linkage Disequilibrium

bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Most regulatory interactions are not in linkage disequilibrium

Sean Whalen1 and Katherine S. Pollard∗1,2,3

1Gladstone Institutes, San Francisco, CA 94158, USA 2Department of Epidemiology and Biostatistics, Institute for Human Genetics, Quantitative Biology Institute, and Institute for Computational Health Sciences, University of California San Francisco, San Francisco, CA, USA 3Chan-Zuckerberg Biohub, San Francisco, CA, USA

February 26, 2018

1 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Abstract

Linkage disequilibrium (LD) and genomic proximity are commonly used to map non-coding variants to genes, despite increasing examples of causal variants outside the LD block of the gene they regulate. We compared chromatin contacts in 22 cell types to LD across billions of pairs of loci in the human genome and found no concordance, even at genomic distances below 25 kilobases where both tend to be high. Gene expression and ontology data suggest that chromatin contacts identify regulatory variants more reliably than do LD and genomic proximity. We conclude that the genomic architectures of genetic and physical interactions are independent, with important implications for gene regulatory evolution and precision medicine.

2 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Genetic variants ranging from large scale chromosomal rearrangements to single nucleotide polymorphisms (SNPs) can impact gene function by altering exonic sequence or by changing gene regulation. Recent studies estimate that 93% of disease-associated variants are in non-coding DNA [1] and 60% of causal variants map to regulatory elements [2], accounting for 79% of phenotypic variance [3]. Additionally, disease- associated variants are enriched in regulatory regions [4], especially those from tissues relevant to the phenotype [5]. Functionally annotating non-coding variants and correctly mapping causal variants to the genes and pathways they affect is critical for understanding disease mechanisms and using genetics in precision medicine [6–9]. Common practice associates non-coding variants with the closest gene promoter or promoters within the same LD block. However, regulatory variants can affect phenotypes by changing the expression of target genes up to several megabases (mb) away [10–13], well beyond their LD block (median length ≈ 1-2kb, Supplementary Table 1b). This prompted Corradin and colleagues to conclude that a gene’s regulatory program is not related to local haplotype structure [14]. Even when a GWAS SNP is in LD with a gene that has a strong biological link to the phenotype, the causal variant may be in a nearby non-coding region regulating a different gene [15, 16]. Highlighting the long range of regulatory interactions, recent work in T cells found that only 14% of 684 autoimmune variants targeted their closest gene; 86% skipped one or more intervening genes to reach their target, and 64% of variants interacted with multiple genes [17]. Thus, many non-coding variants are far away and in low LD with the promoters they regulate. Distal non-coding variants can cause changes in gene regulation and phenotypes via three-dimensional (3D) chromatin interactions. For example, the obesity-associated FTO variant (rs1421085) was found to disrupt an ARID5B repressor motif in an enhancer for IRX3/5 during adipocyte differentiation, increasing obesity risk [10]. A second study showed a schizophrenia-associated SNP (rs1191551) regulates the expression of distal gene FOXG1 in two zones of the developing human cerebral cortex, rather than targeting the nearby gene PRKD1 [13]. Another example is a papillary thyroid cancer associated SNP (rs965513) in an LD block containing several enhancer variants that contact the promoter of FOXE1 and alter its expression [18]. In addition, mutagenesis screens identified multiple distal variants that lead to cancer drug resistance by decreasing CUL3 expression [19]. These validated causal SNPs demonstrate that regulatory variants can be located far from their target promoters in distinct LD blocks (IRX3/5 1.2mb, FOXG1 760kb, CUL3 ± 100kb, FOXE1 ± 60kb). New understanding of the 3D genome from high-throughput chromatin capture (Hi-C) and imaging data suggests it may be common for regulatory variants and their target gene(s) to lack strong LD. For example, mammalian genomes are partitioned into regions enriched for chromatin interactions at multiple scales, including Topologically Associating Domains (TADs, median length 880kb [20]) and contact domains (sub- TADs, median length 250kb [21]). While these chromatin domains resemble the nested block patterns of LD, they have a different origin: insulating chromatin boundary elements across which relatively few chromatin interactions occur versus frequency of recombination events over generations. Furthermore, different proteins interact with DNA to mediate these processes, namely PRDM9 in the case of recombination [22] and structural proteins such as CTCF in the case of chromatin boundaries [23]. Thus, one might not expect similarity a priori. However, LD is high and chromatin interactions are common at genomic distances less than 25 kilobases (kb), so LD and chromatin contact maps might be correlated at this scale even though some causal SNPs regulate promoters over long genomic distances where LD is approximately zero. The correspondence between chromatin contact frequencies and LD has yet to be comprehensively evaluated on a genome-wide scale. To address this question, we quantified the relationship between LD and chromatin contact frequency genome-wide using Hi-C data from 5 diverse cell lines [21] and promoter capture Hi-C (PCHi-C) data from 17 primary blood cell types [12], combined with SNPs from the 1000 Genomes Project [24]. In several analyses, we focus on blood cells to enable integration of B cell eQTLs [25] and blood-relevant data from the GWAS catalog [1] with high resolution chromatin interaction data in a consistent cellular context. Using multiple large scale analyses over billions of chromatin contacts and SNP pairs, we first demonstrate that LD is not a proxy for regulatory interaction. In particular, LD decays at much shorter distances than chromatin contacts and is no higher in regions with statistically significant chromatin interactions than distance matched regions with non-significant interactions. Even at genomic distances < 25kb where LD and contact frequencies are both high, their block patterns are rarely aligned. We then show that utilizing higher-order chromatin organization is essential for accurate functional assignment of non-coding variants, having implications for

3 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

ﬁne mapping of causal variants and applications to precision medicine.

4 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Results

To comprehensively compare the genomic architectures of LD and chromatin contacts, we generated two types of data structures from publicly available data (Supplementary Table 1). The first includes LD blocks and pairwise LD between all high-quality, bi-allelic SNPs across individuals from each of the 1000 Genomes Project superpopulations [24] (AFR: African, AMR: Ad Mixed American, EAS: East Asian, EUR: European, SAS: South Asian). The second records contact frequencies between all pairs of fragments in 22 human cell types with high-resolution Hi-C [21] or promoter capture Hi-C (PCHi-C) [12] data that measures interactions between baited promoters and promoter interacting regions (PIRs). These chromatin contact data were used to generate lists of statistically significant interacting regions and distance-matched regions with non- significant interactions for each cell type using with methods that account for expected contact frequencies and adjust for multiple testing. Significant chromatin interactions from Hi-C were computed using Juicer [26] and represent statistical enrichment of contacts over a particular choice of local background, whereas those from PCHi-C were identified using CHiCAGO [27] and indicate if a region is likely to be in the same contact domain with a promoter or not. Due to the resolution of chromatin interaction assays, we could not compare LD and 3D proximity between sites separated by less than 5kb where both values are expected to be high. The resulting analyses span approximately 1.6 million LD blocks, 27 billion SNP pairs, and 3.1 million statistically significant chromatin interactions (Supplementary Table 1). By analyzing the genome-wide relationship between LD and chromatin contacts from multiple perspectives, we show that LD is not correlated with chromatin interactions and generally should not be used for mapping non-coding SNPs to genes and pathways.

Chromatin interactions and LD have different genomic architectures Both LD and chromatin contact frequency measure the strength of a relationship between pairs of genomic sites. However, these two measures differ fundamentally in their scales: Chromatin contacts span much longer distances (Figure 1). Genetic architecture forms LD blocks of median length 2kb (combined 1000 Genomes super-populations) in which a percentage of SNP pairs exceed a common threshold of R2 > 0.8 [28]. Strong LD pairs have a median distance of 13kb as they can be located in different blocks. On the other hand, physical architecture forms regions enriched for interactions at much longer scales, including contact domains (median 250kb [21]), focal interactions (median 270kb [21] or 350kb [12]), and topological domains (median 840kb). This difference is evident in most genomic loci when contact frequency from a particular cell type is plotted alongside LD from 1000 Genomes, both at the scale of TADs (Figure 2a) and within smaller contact domains (Figure 2b) where chromatin interactions are frequent but LD structure is low or limited to smaller LD blocks. Due to this difference in scale, non-coding SNPs frequently contact genes located hundreds of kb away without being in LD with those genes (Figure 2b). These distal chromatin interactions may differ across cell types, whereas LD does not (Figure 2c). A similar example is shown in Supplementary Figure 1. To quantify the decay rates of LD versus chromatin contacts genome-wide, we analyzed all pairs of sites separated by a given genomic distance with respect to Hi-C contact frequency in ENCODE cell lines [21] versus LD in 1000 Genomes individuals. This showed that contact frequency decays with genomic distance much slower than LD both across (Figure 3, Supplementary Figure 2) and within human populations (Sup- plementary Figure 4). Furthermore, statistically significant chromatin interactions occur between genomic regions separated by dozens, hundreds, or even thousands of LD blocks (Supplementary Figure 6 panels a-b), while most SNP pairs with non-zero LD cross 0-2 contact domains (Supplementary Figure 6, panel c). PCHi-C data shows the same broad trends (Supplementary Figure 3). In summary, genetic and physical architectures of human chromosomes differ at all scales (Figure 1).

Chromatin contact frequencies have low concordance with LD across genomic distances Contact frequency and LD could still be correlated at shorter genomic distances where LD is more often non-zero. To explore this possibility, we analyzed the concordance of frequent Hi-C contacts and strong LD values (R2 > 0.8) for pairs of sites within genomic windows from 5kb to 1.2mb (Methods). Chromatin

5 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

LD Blocks (1KG Phase 3)

Strong LD Pairs (1KG Phase 3, chr 22)

Combined eQTLs (GTEx Consortium 2017)

B cell & Monocyte eQTLs Architecture (Fairfax et al. 2012) Genetic Contact Domains Physical (Rao et al. 2014)

Genomic Feature Chromatin Loops (Rao et al. 2014)

Chromatin Loops (Javierre et al. 2016)

Topological Domains (Dixon et al. 2012) 1e−01 1e+02 1e+05

log(Distance or Size (kilobases))

Figure 1: Genetic and Physical Architectures LD blocks and strong LD pairs (R2 > 0.8) operate across tens of kb or less, while chromatin interactions and multi-scale domains of enrichment span hundreds of kb, with eQTLs roughly in between. Summaries are computed over all super-populations, tissue types, or cell types in the relevant datasets.

interactions and strong LD co-occur about as often as expected if there was no association between the two variables (Figure 4). Concordance is highest at short genomic distances and decays as LD approaches zero. Notably, concordance is nearly zero at scales where statistically signiﬁcant chromatin interactions still occur. The maximum concordance and rate of decay diﬀer across 1000 Genomes super-populations, with the lowest concordance in African populations that have the fastest decay of LD. These patterns are consistent with the two variables being independent and LD decaying more rapidly with genomic distance (Figure 3). Hence, there is no evidence that LD and chromatin contacts are correlated at genomic distances of 5kb or more.

LD is not elevated in significant chromatin interactions Next we compared LD and chromatin structure focusing on statistically significant chromatin interactions, as these might harbor high LD SNPs even if less frequent chromatin contacts are rarely genetically linked. For each statistically significant and distance-matched non-significant interaction, we computed the maximum LD between pairs of SNPs occurring on opposing fragments. The ratio of interacting versus non-interacting fragment LD is often close to 1 across all super-populations and cell types (Figure 5, Supplemental Table 2), indicating no elevation of LD at interacting regions. In addition, LD is very low between non-coding regions and interacting promoters in PCHi-C data, with 2−7% of interacting fragments located within the same LD

6 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

(a) NHEK, EAS, chr4:100475000-107075000

1.0 ) s n i b

8 b k 5 (

0.8 t n

6 u o C

0.6 c a

t Contact Domain 4 n o LD Block C

C Significant Loop - 0.4 i H 2 d e z i l a

0.2 m Mean LD (1kb bins, MAF 0.05) 0 r o N

2 g o 0.0 l

(b) NHEK, EAS, chr4:103475000-104075000 (c) HUVEC, EAS, chr4:103475000-104075000 rs17032996 (eQTL SNP) MANBA (Closest Gene to SNP) SLC9B2 (eQTL Target) rs17032996 (eQTL SNP) MANBA (Closest Gene to SNP) SLC9B2 (eQTL Target)

rs17032996 (eQTL SNP) rs17032996 (eQTL SNP) MANBA (Closest Gene to SNP) MANBA (Closest Gene to SNP)

SLC9B2 (eQTL Target) SLC9B2 (eQTL Target)

Figure 2: Discordance between LD and Hi-C An annotated matrix illustrates differences between the genomic scales of LD [24] (R2, upper triangle, green) versus Hi-C contact frequency [21] (lower triangle, purple). Rows and columns are binned genomic coordinates (hg19) with lower bins near the upper left; for example, row 10 column 11 stores the LD between a bin and its neighbor, while row 11 column 10 stores the contact frequency. More frequent contacts (5kb bins) are darker purple; higher LD (averaged over non-zero LD pairs in 1kb bins) are darker green. Contact domains (nested purple squares) and significant interactions (orange squares) were computed from Hi-C data. LD blocks (green squares) were computed from 1000 Genomes genotypes. While some LD blocks fall within contact domains, there are also many cases where they overlap domain boundaries. (a) A representative 6.6mb locus on chromosome 4 shows Hi-C contacts (NHEK cells) span much longer distances than LD (EAS superpopulation). (b) A 600kb locus on the same chromosome illustrates the complexities of mapping a non-coding SNP (rs17032996) to a target gene. The closest gene MANBA falls within the same LD block as the SNP. However, Hi-C data shows the SNP contacts the SLC9B2 gene ≈ 460kb away in NHEK cells, skipping over intervening expressed gene MANBA. rs17032996 is also an eQTL in B cells [25] and significantly interacts with SLC9B2 in several blood cell types [12]. (c) In HUVEC cells, the SNP no longer interacts with SLC9B2 and several contact domains are lost.

7 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Cell Line Closest Gene LD (AFR) LD (AMR) LD (EAS) LD (EUR) LD (SAS) Mon 9.9% 2.1% 4.3% 5.5% 5.0% 4.8% Mac0 9.4% 1.7% 3.5% 4.4% 4.2% 4.0% Mac1 10.1% 2.1% 4.4% 5.4% 5.1% 4.9% Mac2 10.4% 2.2% 4.6% 5.6% 5.5% 5.2% Neu 9.8% 2.5% 5.1% 6.3% 6.0% 5.4% MK 11.9% 3.0% 5.9% 7.4% 6.9% 6.6% EP 10.9% 2.4% 4.8% 6.0% 5.7% 5.4% Ery 8.7% 2.8% 5.7% 7.0% 6.7% 6.3% FoeT 7.0% 2.0% 4.1% 5.1% 4.8% 4.6% nCD4 8.8% 2.4% 5.0% 6.3% 5.9% 5.6% tCD4 7.9% 2.0% 4.1% 5.2% 4.9% 4.7% aCD4 9.3% 2.3% 4.9% 6.2% 5.8% 5.5% naCD4 8.6% 2.0% 4.3% 5.4% 5.1% 4.8% nCD8 8.4% 2.4% 4.8% 6.1% 5.7% 5.4% tCD8 9.2% 2.7% 5.3% 6.7% 6.2% 5.9% nB 9.0% 2.2% 4.4% 5.8% 5.4% 5.1% tB 8.7% 2.2% 4.2% 5.6% 5.2% 4.9%

Table 1: Accuracy of methods for mapping non-coding regions to genes Using PIRs from statistically signiﬁcant PCHi-C interactions in 17 blood cell types as a gold standard [12], we computed the percent agreement of various heuristics for mapping non-coding regions to all baited promoters genome-wide. The closest gene method predicts that a non-coding region interacts with the nearest promoter. The LD heuristic predicts that a non-coding region interacts with any gene in its LD block (one prediction per 1000 Genomes superpopulation). The agreement of LD is even lower than that of the closest gene: 2-7% versus 7-12% across superpopulations.

block (Table 1). Thus, LD is not an adequate proxy for high-conﬁdence chromatin interactions, including those involving promoters.

LD is low between distal regulatory SNPs and their genes Genetic variants associated with statistically significant differences in a gene’s expression (eQTLs) provide evidence of functional relationships between regulatory regions and genes separated by long genomic distances. Indeed, target genes of GTEx eQTLs [29] and blood eQTLs [25] have median distances of 49 and 113kb, respectively. Combined with our other findings, these distances suggest that a distal eQTL and its target gene are likely to have zero LD and thus be separated by a large number of LD blocks. We therefore compared the frequency of eQTLs amongst non-coding regions that interact with gene promoters versus distance-matched regions that do not. Analyzing B-cells where both PCHi-C [12] and eQTL [25] data were available, statistically significant chromatin interactions were highly enriched for eQTLs across genomic distances up to 1.6mb (Figure 6, Figure 1). In contrast, regions in strong LD with a promoter were only enriched for eQTLs at genomic distances less than 100kb, emphasizing that more distal eQTLs are often in 3D proximity to their target promoters but are not genetically linked to them. This result holds across all super-populations. Thus, genomic distance and LD are poor predictors of the genes targeted by distal eQTLs [11].

Mapping non-coding variants to genes with Hi-C produces more functional enrichments than genomic distance or LD If regulatory interactions are common at large genomic distances where LD is approximately zero, then GWAS SNPs linked to genes via Hi-C should include more true gene targets than using closest genes or genes in LD with the SNP. If true, then the set of genes associated with GWAS hits via Hi-C should also share more functional annotations. To test this idea, we examined the magnitude and statistical signiﬁcance

8 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

of Gene Ontology (GO) enrichments for genes associated with all GWAS SNPs for a given phenotype via PCHi-C interactions [12] (all genes with promoter interacting regions overlapping the SNP), genomic distance (closest promoter to the SNP), or genetic distance (all promoters in the same LD-block as the SNP). Most blood-relevant phenotypes in the GWAS catalog [30] had the largest number of significantly enriched GO terms using blood cell PCHi-C assignment (Figure 7), even after controlling for the number of genes mapped to GWAS SNPs. LD-based assignment occasionally produced a limited number of terms with effect sizes larger than those from PCHi-C assignment (Figure 7c). Nonetheless, using LD resulted in fewer GO terms associated with the phenotype and a lower area-under-the-curve than PCHi-C. As a negative control, we examined GWAS SNPs for phenotypes not relevant to blood and found that closest gene and LD approaches have more GO enrichments than for blood-relevant phenotypes, confirming that PCHi-C enrichments are tissue-specific. Thus, PCHi-C data from blood does not correctly associate SNPs with genes for non-blood phenotypes, highlighting the need for interaction data collected in the cell type of interest to harness the power of chromatin interactions for functional assignment.

9 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

(a) Hi-C (K562) (b) Hi-C (GM12878) (c) Hi-C (IMR90)

3 3 3

2 2 2

1 1 1

(Hi−C Contact Count) 0 (Hi−C Contact Count) 0 (Hi−C Contact Count) 0 10 10 10 g g g o o o l l l −1 −1 −1 0 0 0 25 50 75 25 50 75 25 50 75 100 100 100 Genomic Distance (kilobases) Genomic Distance (kilobases) Genomic Distance (kilobases)

(d) Hi-C (NHEK) (e) Hi-C (HUVEC) (f) LD (All)

3 3 1.00

2 2 0.75 )

2 R 1 1 0.50 LD (

(Hi−C Contact Count) 0 (Hi−C Contact Count) 0 0.25 10 10 g g o o l l −1 −1 0.00 0 0 0 25 50 75 25 50 75 25 50 75 100 100 100 Genomic Distance (kilobases) Genomic Distance (kilobases) Genomic Distance (kilobases)

Figure 3: LD and Hi-C contact decay with genomic distance Both Hi-C contact frequency [21] (panels a-e) and LD (panel f) are anti-correlated with genomic distance (Spearman ρ between −0.5 and −0.71 for Hi-C across cell lines; ρ ≈ −0.52 for LD). All plots display non-zero values from their respective datasets. LD decays towards zero at much shorter genomic distance than contact frequency, with high LD SNP pairs concentrated below 50kb. In contrast, Hi-C contacts are common up to and exceeding the median length of contact domains (250kb) or TADs (840kb). Supplementary Figure 2 highlights decay up to 2mb, while this ﬁgure highlights decay up to 100kb. Supplementary Figure 4 shows nearly identical LD scaling per superpopulation. Contact frequencies vary in approximate proportion to sequencing depth and number of replicates per cell line (Supplementary Table 1). Panel f) shows 836 million biallelic SNPs on chromosome 14 and is representative of other chromosomes.

10 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Observed Expected

AFR AMR EAS EUR SAS 50 40

● K562

● ● ● 30 ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 20 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 10 ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● 0 ● ● ● 50

40 GM12878

●

● ● ●

● ● 30 ● ● ● ● ● ● ● ●

● ● ●

● ● ● ●

● ●

● ● 20 ●

● ●

●

● 10 ●

● ● ●

● ● ● ● ● ● ● ● ● 0 ● ● 50 40

● IMR90

● ● ●

● 30 ● ● ● ● ● ● ● ●

● ● ●

● ● ● ● ●

● ●

● ● ● 20 ● ● ● ● ● ● ● ● ● ●

● ● ● 10 ● ● ● ● ●

● ● ● ● ● ● ● 0 ● ● 50 40

● NHEK

● ● ●

● ● ● 30 ● ● ● ● ● ● ●

● ● ● ●

● ● ● ● ●

● ●

● ● 20 ● ● ●

● ● ● ●

● ●

● ● ● 10 ● ● ● ● ●

● ● ● ● ● ● ● 0 ● ● ●

Strong LD & Contact Frequency Co−Occurrence (Percent) 50 40 HUVEC

●

● ● ●

● 30 ● ● ● ● ● ● ● ● ●

● ● ●

● ● ● ● ● ● ● ●

● ● 20 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 10 ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0 ● ● ● 5 5 5 5 5 10 20 40 80 10 20 40 80 10 20 40 80 10 20 40 80 10 20 40 80 160 320 640 160 320 640 160 320 640 160 320 640 160 320 640 1280 1280 1280 1280 1280 Window Size (kilobases)

Figure 4: Genome-wide Hi-C and LD concordance Frequent Hi-C contacts and strong LD (R2 > 0.8) co-occur less than 50% of the time at short genomic distances. Concordance is cut nearly in half at 40kb where most LD has decayed to 0, and is nearly 0 at many scales where statistically signiﬁcant chromatin interations occur. In addition, concordance and rate of decay varies by super-population, with AFR having only ≈ 12% concordance at short genomic distances

11 bioRxiv preprint not certifiedbypeerreview)istheauthor/funder,whohasgrantedbioRxivalicensetodisplaypreprintinperpetuity.Itmadeavailable otiigtecoetgn r nihdfreTsa hre itnebn pnlb,wieinteracting while b), (panel bins distance c). shorter (panel bin at proximal fragments eQTLs most bait pairwise the for (maximum with at enriched LD fragments strong are Promoter-interacting in gene fragments a). closest (panel the [12] containing interactions PCHi-C B-cell significant 6: Figure LD. weak in a differences not ratios small is figure; relatively LD this of that for result indicating values the superpopulations, raw are and provides in 1 types 2 SNPs than Table cell for larger Supplemental all enriched or for interactions. are smaller 1 chromatin interactions near for significant is proxy proxy indicates of a ratio sufficient 1 ratio as this than The acts However, greater LD datasets. LD. well ratio PCHi-C strong how 17 a quantifies and interactions; interactions Hi-C non-significant chromatin versus 5 for significant for for computed LD was maximum interaction LD) the mean (interaction interactions, SNPs chromatin between non-significant LD distance-matched pairwise and significant statistically of ments 5: Figure Odds Ratio for eQTL Enrichment Superpopulation AMR EUR SAS EAS AFR 10 20 30 40 0.8 0.8 0.8 0.9 0.8 doi: QLenrichment eQTL infiatvru o-infiatcrmtnitrcinLD interaction chromatin non-significant versus Significant 0−.13 K562 https://doi.org/10.1101/272245 0.8 0.9 0.9 0.9 0.9 Chromatin Interactions Statistically Significant GM12878 0.8 .13−.22 0.8 0.8 0.8 0.9 IMR90 0.8 0.8 0.8 0.9 0.8 (a) .22−.35 NHEK 0.8 0.8 0.8 0.9 0.8 HUVEC .35−.57 1.0 1.0 1.0 1.0 1.0

-eleTs[5 r nihdars uniebne itne nstatistically in distances quantile-binned across enriched are [25] eQTLs B-cell Mon 0.9 0.9 0.9 0.9 0.9 .57−1.6 Mac0 under a 0.9 0.9 0.9 0.9 0.9 ;

Mac1 this versionpostedFebruary26,2018. CC-BY-NC-ND 4.0Internationallicense 0 2 4 6 1.0 1.0 1.0 1.0 1.0 Genomic Distance(megabases) p−value <0.01 Mac2 R 0−.13 1.1 1.1 1.1 1.1 1.1 2 Neu Inside BaitFragments > Cell Type 1.0 1.0 1.0 1.0 1.0 ,aeae cossprouain)aeol enriched only are superpopulations) across averaged 0.8, .13−.22 Closest Gene MK 12 1.0 1.0 1.0 0.9 1.0 EP (b) ● 1.1 1.1 1.1 1.1 1.1

.22−.35 ● Ery FALSE 1.0 1.0 1.1 1.0 1.0 .35−.57 ● FoeT 1.0 1.0 1.0 1.0 1.0 nCD4 ●

.57−1.6 TRUE 1.0 1.0 1.0 1.0 1.0

tCD4 The copyrightholderforthispreprint(whichwas . 1.0 1.0 1.0 1.0 1.0

−1 aCD4 0 1 2 3 4 1.0 1.0 1.0 1.0 1.0 o Nslctdo h frag- the on located SNPs For naCD4

0−.13 1.0 1.0 1.1 1.0 1.0 nCD8 Between Fragments 1.0 1.0 1.0 1.0 1.0 .13−.22 ● tCD8 Strong LD 1.0 1.0 1.0 1.0 1.0

.22−.35 1.0 1.0 1.0 1.0 1.0 tB

.35−.57 ● ●

.57−1.6 0.0 0.4 0.8 1.2 1.6 2.0 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

(a) Diastolic Blood Pressure (b) Mean Platelet Volume

Closest Gene LD Block PCHi−C Closest Gene LD Block PCHi−C 12.5 10.0

10.0

7.5 7.5

5.0 5.0

−log10(q−value) 2.5 −log10(q−value) 2.5

0.0 0.0 GO Term (sorted by q−value) GO Term (sorted by q−value)

Closest Gene LD Block PCHi−C Closest Gene LD Block PCHi−C

2.0 3

1.5

2 1.0

−log10(q−value) −log10(q−value) 1 0.5

0.0 0 GO Term (sorted by q−value) GO Term (sorted by q−value)

Figure 7: GO enrichment for closest, same-LD-block, and PCHi-C bait genes Enrichment of GO terms (Benjamini-Hochberg adjusted p-values, -log10 scale) in multiple blood-relevant phenotypes from the GWAS catalog [30]. Methods for functional assignment of SNPs include using the closest gene, all genes within the same LD block (EUR super-population), and promoter capture bait genes with a SNP located in the promoter-interacting region of a statistically significant blood cell chromatin interaction [12]. Gray horizontal lines indicate FDR cutoffs of 1, 5, and 10 percent. In blood-relevant phenotypes (panels a-c), PCHi-C bait genes interacting in 3D with GWAS SNPs typically show more enrichment for a larger number of terms than same-LD-block or closest gene approaches, whose enrichment is affected by large numbers of false positives and negatives. For non-blood phenotypes, chromatin interactions in the wrong cell type can have little or no enrichment compared to closest gene and same-LD-block approaches (panel d).

13 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Discussion

Chromatin interactions and LD are both pairwise measurements between genomic loci that show block patterns along mammalian chromosomes. Given their similar structure, it might be tempting to speculate that LD blocks correspond to or are contained within three-dimensional chromatin domains. In this study, we conducted diverse genome-wide analyses across different length scales and showed that LD maps are not correlated with chromatin interaction maps. The main factor driving these differences is the frequency of chromatin interactions over genomic distances where the genetic linkage between SNPs is close to zero. Significant chromatin interactions often span hundreds or sometimes thousands of LD blocks. However, even at genomic distances where contact frequencies and LD are both high on average, the correlation between their block patterns is weak. This discordance is perhaps not surprising given the different origins of chromatin domains (insulated regulatory regions) and LD blocks (recombination), as well as the increasing number of examples of distal enhancers and eQTLs documented in the literature. Despite growing awareness that regulatory interactions need not be in LD or nearby on the genome, this phenomenon appears not to have been quantified systematically genome-wide until now. To do so, we developed computationally efficient statistical analyses and leveraged 5 ENCODE cell lines [21], 17 human primary blood cell types [12], and 5 superpopulations from the 1000 Genomes Project [24]. The results document that LD and chromatin interactions are indeed uncorrelated at genomic distances of 5kb or more. This study has important implications for associating non-coding variants with genes, downstream phenotypes, and molecular mechanisms. Our results show that variants have great potential to regulate genes beyond their LD block. Hence, the common practice of mapping candidate regulatory variants to the closest gene or other genes in the same LD block will typically miss most chromatin interactions between the variant and gene promoters. In addition, LD is the same in all cell types whereas within-TAD regulatory interactions vary across cell types [21] (Figure 2, Supplementary Figure 1). Despite the emergence of chromatin interaction data as a new paradigm for mapping non-coding variants to genes, LD blocks and genomic distance are still widely used, resulting in many GWAS hits being annotated with incorrect genes. There are good reasons why genomic distance and LD have been used to annotate regulatory variants: these quantities are relatively easy to compute, and only a few cell types currently have high enough resolution chromatin interaction data (≈ 1-5kb) for linking specific regulatory variants to promoters. The lack of correlation between these common heuristics and high resolution chromatin contacts underscores the importance of generating or computationally predicting chromatin structure across many more cell types. In addition to highlighting the need for incorporating chromatin interactions into functional assignment, the discordance between chromatin contact frequency and LD has evolutionary implications. One consequence is that entire TADs or sub-TADs do not typically segregate as single haplotypes in human populations, enabling independent selection on regulatory variants versus the promoter and coding variants of their target genes. Furthermore, recombination blocks that span chromatin domain boundaries indicate that regulatory and coding variants from one domain can segregate with variants from the adjacent domain. The fact that haplotype breakpoints do not align with chromatin boundaries may indicate that recombination is deleterious at these functional elements, perhaps due to the mutagenic effects of recombination. These findings are different from observations regarding fixed structural differences between genomes of different mammals, which tend to preserve TADs with breakpoints enriched at TAD boundaries [31, 32]. We therefore conclude that while chromatin domains are functional genomic entities maintained as syntenic units over evolutionary time, recombination is independent of chromatin structure. This creates novel haplotypes of the genomic elements within and between TADs on which selection can operate.

14 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Methods

In order to perform large scale analyses, some caveats were necessary in order to place reasonable bounds on compute time and memory, even in a high-performance computing environment. For example, LD was computed between SNP pairs at most 2mb apart and stored if LD was 0.01 or greater. Also, the resolution of Hi-C and PCHi-C data prevented examining correlations between chromatin interactions and LD at genomic distances below 5kb. To make our choices transparent and our analyses reproducible, our code is available at https://github.com/shwhalen/loopdis.git. Hi-C data [21] was obtained from the NCBI Gene Expression Omnibus using accession GSE63525 (including contact domains, statistically signiﬁcant loops, and sparse contact matrices along with coeﬃcients for normalization and expectation). Promoter capture Hi-C data [12] was obtained from Open Science Frame- work (https://osf.io/u8tzp/). All data sources use human reference genome hg19. Analyses utilized bcftools 1.6, bedtools 2.27.1 [33], plink 1.90b5 [34], pandas 0.22.0 [35], matplotlib 2.1.1 [36], seaborn 0.8.1, ggplot 2.2.1 [37], and GNU Parallel 20171222 [38]. Python 3.6.4 was provided by the Miniconda distribution; R 3.4.3 was compiled from source using gcc 7.2.1.

Linkage Disequilibrium Bi-allelic SNPs from phase 3 of the 1000 Genome Project were first converted to plink’s binary BED format (--make-bed --allow-extra-chr --biallelic-only), and the pairwise LD computed (--r2) for all SNPs with a minimum MAF of 5% (--maf 0.05) located within 2mb of each other (--ld-window-kb 2000). The number of pairwise comparisons allowed within a window was increased (--ld-window 10000), and the default R2 filter lowered from 20% down to 1% (--ld-window-r2 0.01). Pairs below this threshold were assigned R2 = 0. LD computations were performed separately for each superpopulation by using the 1000 Genomes panel file (integrated call samples v3.20130502.ALL.panel) and the --filter flag. LD blocks were computed using plink (utilizing the algorithm from Gabriel et al. [28]) with the --blocks no-pheno-req no-small-max-span --blocks-max-kb 2000 flags. Blocks were computed separately for each superpopulation using the same --filter flag and panel file.

Interacting versus Non-Interacting LD For each 1000 Genomes superpopulation, bi-allelic SNPs with a minimum MAF of 5% were intersected with either Hi-C [21] or promoter capture Hi-C [12] fragments using the bedtools pairtobed command with the -type both flag. For each pair of interacting fragments, the maximum LD between SNPs on different fragments was computed. The mean of this maximum pairwise LD was computed separately for statistically significant and non-significant interactions in order to compute a ratio. For Hi-C data, negatives (non-significantly interacting fragments) were obtained by shuffling a list of positives (significant interactions) called by the Juicer pipeline at 10% FDR [26] along the same chromosome. For promoter capture Hi-C data, positives and negatives were obtained from a list of interactions scored by the CHiCAGO pipeline [27]. As in the original paper, negatives were interactions with a score less than 5. For both datasets, negatives were distance-matched to positives using quantile binning of interaction distance.

Hi-C versus LD Concordance Observed over expected Hi-C values were computed using formulas from Rao et al. [21] applied to VC- normalized contact counts at 5kb resolution for each ENCODE cell line. For comparable resolution, LD per 5kb genomic bin was computed for each 1000 Genomes superpopulation using the 75th percentile of pairwise LD values in the bin. This was more robust outliers and heavily zero-skewed LD distributions than the average or median. Concordance was computed based on whether a bin’s LD value was strong (R2 > 0.8) and its chromatin contact frequency was strong (above the 75th percentile of contact frequencies), for all bins located in non- overlapping genomic windows of ﬁxed size. This was repeated for window sizes of 5, 10, 20, 40, 80, 160, 320, 640, and 1280kb to examine concordance across multiple scales, and without variation introduced by diﬀerent TAD-calling algorithms.

15 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

eQTL Statistics B-cell eQTLs [25] were intersected with naive B-cell promoter capture Hi-C interactions [12]; the eQTL was required to overlap the promoter-interacting region and the eQTL target was required to overlap the bait fragment. The presence or absence of an interacting eQTL was stored in a binary vector. Next, the closest gene to each promoter-interacting fragment was computed using bedtools closest and Ensembl gene annotations. The presence or absence of the closest gene in the corresponding bait fragment was stored in a binary vector. Next, the statistical significance of chromatin interactions (thresholded using a CHiCAGO score of 5) was stored in a binary vector. Finally, for each superpopulation, a binary vector stored the maximum pairwise LD between fragments. eQTLs were tested for enrichment in statistically significant chromatin interactions, interactions where the bait was the closest gene, and interactions where the maximum pairwise LD between fragments was > 0.8 (averaged over superpopulations). Interactions were quantile binned by distance up to 1.6mb so that all entries in the contingency table would be non-zero. For each distance bin, the odds ratio was computed as the ratio of the diagonal to the off-diagonal entries of the contingency table, and the p-value was computed using R’s fisher.test function.

Gene Ontology Enrichment The promoter-interacting region of statistically signiﬁcant PCHi-C interactions [12] was intersected with SNPs for the 30 most abundant phenotypes in the GWAS catalog [30] (release 2018-01-31). For each GO term, a Fisher’s exact test was computed on a 2 by 2 contingency table counting if the interaction contained a GWAS SNP for the phenotype in its PIR, and whether or not the interaction’s bait gene was annotated with that GO term. Benjamini-Hochberg correction for multiple hypothesis testing was applied to the resulting p-values. For comparison, this was repeated for the closest gene to each GWAS SNP, as well as all genes in the same LD block as the GWAS SNP.

16 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Acknowledgments

Thanks to Dr. Hunter Fraser and Dr. Jonathan Pritchard for GO analysis suggestions. Dr. Noah Zaitlen, Dr. Dan Geschwind, Dr. Hyejung Won, and Dr. Marisa Wong Medina provided helpful feedback. This project was supported by the Bench to Bassinet Program of the NHLBI (U01HL098179, UM1HL098179), NIH/NHLBI (HL089707), NIH/NIMH (MH109907), the San Simeon Fund, and the Gladstone Institutes.

17 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Author contributions

SW and KSP designed the experiments; SW coded the experiments; SW and KSP wrote the paper.

18 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Competing interests

The authors declare that they have no competing interests.

19 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Supplemental Material

20 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Supplemental Tables

(a) Cell Type Statistics Assay Cell Type Acronym Reps. Unique Read Pairs Sig. Interactions Hi-C Lymphoblastoid cells GM12878 9 4,907,147,001 9,448 Hi-C Fetal lung ﬁbroblasts IMR90 2 1,136,673,290 8,040 Hi-C Epidermal skin keratinocytes NHEK 1 664,899,299 4,929 Hi-C Erythroleukemia cells K562 2 932,208,867 6,057 Hi-C Umbilical vein endothelial cells HUVEC 1 460,393,495 3,865 PCHi-C Megakaryocytes MK 4 653,848,788 150,779 PCHi-C Erythroblasts Ery 3 588,786,672 151,215 PCHi-C Neutrophils Neu 3 736,055,569 142,435 PCHi-C Monocytes Mon 3 572,357,387 165,947 PCHi-C Macrophages M0 Mac0 3 668,675,248 180,190 PCHi-C Macrophages M1 Mac1 3 497,683,496 171,031 PCHi-C Macrophages M2 Mac2 3 523,561,551 186,172 PCHi-C Endothelial precursors EP 3 420,536,621 145,888 PCHi-C Naive B cells nB 3 629,928,642 189,720 PCHi-C Total B cells tB 3 702,533,922 213,539 PCHi-C Fetal thymus FoeT 3 776,491,344 166,743 PCHi-C Naive CD4+ T cells nCD4 4 844,697,853 210,074 PCHi-C Total CD4+ T cells tCD4 3 836,974,777 199,525 PCHi-C Non-activated total CD4+ T cells naCD4 3 721,030,702 211,720 PCHi-C Activated total CD4+ T cells aCD4 3 749,720,649 213,235 PCHi-C Naive CD8+ T cells nCD8 3 747,834,572 216,232 PCHi-C Total CD8+ T cells tCD8 3 628,771,947 204,382

(b) LD Statistics Superpop. # SNP Pairs # LD Blocks BL .05 BL .5 BL .95 BP .05 BP .5 BP .95 AFR 5,447,813,153 565,469 10 1,106 16,540 2 5 53 AMR 10,353,472,856 296,861 9 2,076 33,784 2 7 81 EAS 3,166,821,307 216,961 8 2,833 47,733 2 8 105 EUR 4,156,780,073 242,446 8 2,502 43,067 2 7 100 SAS 4,079,746,085 266,508 9 2,277 38,378 2 7 93

Supplementary Table 1: Dataset statistics a) Summary of Hi-C [21] and promoter capture Hi-C (PCHi-C) [12] datasets. Statistically significant (sig.) Hi-C interactions were called on combined replicates (reps.) by the Juicer pipeline with a 10% false discovery rate (FDR) [26], while PCHi-C interactions were scored by the CHiCAGO pipeline [27] with a significance threshold of 5. PCHi-C has 15-17 fold enrichment for promoter interactions, resulting in an effective coverage of 165 billion read pairs compared to 15 billion for Hi-C [12]. b) Number of unique SNP pairs and LD blocks per superpopulation. LD block quantiles (.05, .5, and .95) for length (BL) and unique SNP pairs (BP) are also given. SNPs were filtered as described in Methods.

21 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

mean(max(R2)) Cell Type Superpopulation Signiﬁcant K562 AFR 0 0.193 1 0.163 AMR 0 0.242 1 0.216 EAS 0 0.163 1 0.137 EUR 0 0.191 1 0.162 SAS 0 0.172 1 0.144 GM12878 AFR 0 0.161 1 0.141 AMR 0 0.202 1 0.183 EAS 0 0.145 1 0.123 EUR 0 0.167 1 0.145 SAS 0 0.152 1 0.129 IMR90 AFR 0 0.198 1 0.165 AMR 0 0.242 1 0.209 EAS 0 0.177 1 0.145 EUR 0 0.203 1 0.169 SAS 0 0.186 1 0.153 NHEK AFR 0 0.216 1 0.175 AMR 0 0.267 1 0.233 EAS 0 0.167 1 0.133 EUR 0 0.200 1 0.166 SAS 0 0.177 1 0.142 HUVEC AFR 0 0.209 1 0.176 AMR 0 0.260 1 0.230 EAS 0 0.166 1 0.139 EUR 0 0.199 1 0.169 SAS 0 0.178 Continued on next page

22 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

mean(max(R2)) Cell Type Superpopulation Signiﬁcant 1 0.149 Mon AFR 0 0.151 1 0.145 AMR 0 0.192 1 0.183 EAS 0 0.149 1 0.145 EUR 0 0.164 1 0.158 SAS 0 0.152 1 0.146 Mac0 AFR 0 0.142 1 0.124 AMR 0 0.183 1 0.160 EAS 0 0.139 1 0.122 EUR 0 0.155 1 0.136 SAS 0 0.143 1 0.124 Mac1 AFR 0 0.147 1 0.134 AMR 0 0.188 1 0.171 EAS 0 0.144 1 0.133 EUR 0 0.160 1 0.146 SAS 0 0.148 1 0.136 Mac2 AFR 0 0.146 1 0.141 AMR 0 0.187 1 0.178 EAS 0 0.144 1 0.139 EUR 0 0.159 1 0.154 SAS 0 0.147 1 0.142 Neu AFR 0 0.157 1 0.174 AMR 0 0.198 1 0.212 EAS 0 0.155 1 0.170 EUR 0 0.171 1 0.184 Continued on next page

23 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

mean(max(R2)) Cell Type Superpopulation Signiﬁcant SAS 0 0.159 1 0.171 MK AFR 0 0.160 1 0.162 AMR 0 0.202 1 0.200 EAS 0 0.159 1 0.163 EUR 0 0.175 1 0.177 SAS 0 0.162 1 0.165 EP AFR 0 0.148 1 0.141 AMR 0 0.190 1 0.177 EAS 0 0.145 1 0.140 EUR 0 0.161 1 0.154 SAS 0 0.148 1 0.144 Ery AFR 0 0.148 1 0.162 AMR 0 0.189 1 0.199 EAS 0 0.146 1 0.161 EUR 0 0.162 1 0.177 SAS 0 0.149 1 0.163 FoeT AFR 0 0.128 1 0.131 AMR 0 0.168 1 0.169 EAS 0 0.125 1 0.132 EUR 0 0.139 1 0.144 SAS 0 0.128 1 0.133 nCD4 AFR 0 0.149 1 0.151 AMR 0 0.191 1 0.190 EAS 0 0.148 1 0.153 EUR 0 0.163 Continued on next page

24 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

mean(max(R2)) Cell Type Superpopulation Signiﬁcant 1 0.166 SAS 0 0.151 1 0.155 tCD4 AFR 0 0.139 1 0.141 AMR 0 0.179 1 0.180 EAS 0 0.137 1 0.143 EUR 0 0.151 1 0.156 SAS 0 0.140 1 0.144 aCD4 AFR 0 0.151 1 0.151 AMR 0 0.192 1 0.191 EAS 0 0.149 1 0.154 EUR 0 0.165 1 0.168 SAS 0 0.152 1 0.155 naCD4 AFR 0 0.145 1 0.145 AMR 0 0.187 1 0.184 EAS 0 0.144 1 0.147 EUR 0 0.159 1 0.161 SAS 0 0.147 1 0.148 nCD8 AFR 0 0.146 1 0.151 AMR 0 0.187 1 0.190 EAS 0 0.144 1 0.153 EUR 0 0.159 1 0.165 SAS 0 0.147 1 0.154 tCD8 AFR 0 0.153 1 0.155 AMR 0 0.195 1 0.194 EAS 0 0.153 1 0.160 Continued on next page

25 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

mean(max(R2)) Cell Type Superpopulation Significant EUR 0 0.168 1 0.172 SAS 0 0.156 1 0.160 nB AFR 0 0.145 1 0.142 AMR 0 0.187 1 0.180 EAS 0 0.144 1 0.144 EUR 0 0.159 1 0.157 SAS 0 0.146 1 0.145 tB AFR 0 0.143 1 0.141 AMR 0 0.184 1 0.178 EAS 0 0.141 1 0.143 EUR 0 0.156 1 0.155 SAS 0 0.144 1 0.144 Supplementary Table 2: Significant versus non-significant interaction LD For SNPs located on the fragments of statistically significant and distance-matched non-significant chromatin interactions, the maximum pairwise LD between SNPs (interaction LD) was computed for 5 Hi-C and 17 PCHi-C datasets. The mean interaction LD per cell type is given for statistically significant (1) and non-significant (0) interactions. These means were used to compute the ratios shown in Figure 5.

26 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Supplemental Figures

27 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

(a) GM12878, EAS, chr14:85440381-93881268

1.0 12

10 ) s n

0.8 i b

b k

8 5 (

t n u o C

0.6 6 t c a

t Contact Domain n

o LD Block C

C Significant Loop

4 - i H

0.4 d e z i l

2 a m Mean LD (1kb bins, MAF 0.05) r o N

0.2 2 0 g o l

2 0.0

(b) GM12878, EAS, chr14:88440381-88881268 (c) NHEK, EAS, chr14:88440381-88881268 rs73312867 GPR65 KCNK10 SPATA7 rs73312867 GPR65 KCNK10 SPATA7

rs73312867 rs73312867 GPR65 GPR65

KCNK10 KCNK10

SPATA7 SPATA7

Supplementary Figure 1: Discordance between LD and Hi-C An annotated matrix illustrates diﬀerences between the genomic scales of LD [24] (R2, upper triangle, green) versus Hi-C contact frequency [21] (lower triangle, purple). Rows and columns are binned genomic coordinates (hg19) with lower bins near the upper left; for example, row 10 column 11 stores the LD between a bin and its neighbor, while row 11 column 10 stores the contact frequency. More frequent contacts (5kb bins) are darker purple; higher LD (averaged over non-zero LD pairs in 1kb bins) are darker green. Contact domains (nested purple squares) and signiﬁcant interactions (orange squares) were computed from Hi-C data. (a) A representative 8.5mb locus on chromosome 14 shows Hi-C contacts (GM12878 cells) span much longer distances than LD (EAS superpopulation). (b) A 400kb locus on the same chromosome illustrates the complexities of mapping a non-coding SNP (rs73312867) to a target gene. The closest gene GPR65 falls within the same LD block as the SNP. However, Hi-C data shows the SNP contacts the SPATA7 gene ≈ 380kb away, skipping over intervening gene KCNK10. (c) In NHEK cells, the SNP interacts with KCNK10 instead.

28 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

(a) Hi-C (K562) (b) Hi-C (GM12878) (c) Hi-C (IMR90)

3 3 3

2 2 2

1 1 1

0 0 0 (Hi−C Contact Count) (Hi−C Contact Count) (Hi−C Contact Count) 10 10 10 g g g o o o l l l −1 −1 −1 0 0 0 500 500 500 1000 1500 2000 1000 1500 2000 1000 1500 2000 Genomic Distance (kilobases) Genomic Distance (kilobases) Genomic Distance (kilobases)

(d) Hi-C (NHEK) (e) Hi-C (HUVEC) (f) LD (All)

3 3 1.00

2 2 0.75 )

2 R 1 1 0.50 LD (

0 0 0.25 (Hi−C Contact Count) (Hi−C Contact Count) 10 10 g g o o l l −1 −1 0.00 0 0 0 500 500 500 1000 1500 2000 1000 1500 2000 1000 1500 2000 Genomic Distance (kilobases) Genomic Distance (kilobases) Genomic Distance (kilobases)

Supplementary Figure 2: LD and Hi-C contacts decay with genomic distance Both Hi-C contact frequency [21] (panels a-e) and LD (panel f) are anti-correlated with genomic distance (Spearman ρ between −0.5 and −0.71 for Hi-C across cell lines; ρ ≈ −0.52 for LD). All plots display non-zero values from their respective datasets. LD decays towards zero at much shorter genomic distance than contact frequency, with most high LD SNP pairs concentrated below 50kb. Hi-C contacts are common at longer genomic distances up to and exceeding the median length of contact domains (250kb) or TADs (840kb). Figure 3 highlights decay up to 100kb, while this ﬁgure highlights decay up to 2mb. Supplementary Figure 4 shows nearly identical LD scaling per superpopulation. Panel f) shows 836 million biallelic SNPs on chromosome 14 and is representative of other chromosomes.

29 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

(a) PCHi-C (aCD4) (b) PCHi-C (EP) (c) PCHi-C (Ery) (d) PCHi-C (FoeT)

10.0 10.0 10.0 10.0

1.0 1.0 1.0 1.0 (CHiCAGO Score) (CHiCAGO Score) (CHiCAGO Score) (CHiCAGO Score) (CHiCAGO 10 10 10 10

g 0.1 g 0.1 g 0.1 g 0.1 o o o o l l l l 0 0 0 0 500 500 500 500 1000 1500 2000 1000 1500 2000 1000 1500 2000 1000 1500 2000 Genomic Distance (kilobases) Genomic Distance (kilobases) Genomic Distance (kilobases) Genomic Distance (kilobases)

(e) PCHi-C (Mac0) (f) PCHi-C (Mac1) (g) PCHi-C (Mac2) (h) PCHi-C (MK)

10.0 10.0 10.0 10.0

1.0 1.0 1.0 1.0 (CHiCAGO Score) (CHiCAGO Score) (CHiCAGO Score) (CHiCAGO Score) (CHiCAGO 10 10 10 10

(i) PCHi-C (Mon) (j) PCHi-C (naCD4) (k) PCHi-C (nB) (l) PCHi-C (nCD4)

10.0 10.0 10.0 10.0

1.0 1.0 1.0 1.0 (CHiCAGO Score) (CHiCAGO Score) (CHiCAGO Score) (CHiCAGO Score) (CHiCAGO 10 10 10 10

Supplementary Figure 3: PCHi-C contact decay with genomic distance Promoter Capture Hi-C contacts are anti-correlated with distance in 17 blood cell types. Decay is similar to Supplementary Figure 2, though here an interaction score (quantified by CHiCAGO [27]) is used rather than normalized contact frequency. A score above 5 denotes statistically significant interactions, and this threshold is indicated with a dashed line. Interaction distances are concentrated below 1mb, though a large number of significant interactions exist at 500kb, well beyond the 40kb dropoff in LD seen in Figure 3.

30 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

(m) PCHi-C (nCD8) (n) PCHi-C (Neu) (o) PCHi-C (tB) (p) PCHi-C (tCD4)

10.0 10.0 10.0 10.0

1.0 1.0 1.0 1.0 (CHiCAGO Score) (CHiCAGO Score) (CHiCAGO Score) (CHiCAGO Score) (CHiCAGO 10 10 10 10

(q) PCHi-C (tCD8)

10.0

1.0 (CHiCAGO Score) (CHiCAGO 10

g 0.1 o l 0 500 1000 1500 2000 Genomic Distance (kilobases)

Supplementary Figure 3: PCHi-C contact decay with genomic distance (cont.) Promoter Capture Hi-C contacts are anti-correlated with distance in 17 blood cell types. Decay is similar to Supplementary Figure 2, though here an interaction score (quantified by CHiCAGO [27]) is used rather than normalized contact frequency. A score above 5 denotes statistically significant interactions, and this threshold is indicated with a dashed line. Interaction distances are concentrated below 1mb, though a large number of significant interactions exist at 500kb, well beyond the 40kb dropoff in LD seen in Figure 3.

31 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

AFR (ρ = − 0.51) AMR (ρ = − 0.32) EAS (ρ = − 0.62) 1.00

0.75

0.50

0.25

0.00 )

2 0 R 25 50 75

100 EUR (ρ = − 0.58) SAS (ρ = − 0.6)

LD ( 1.00

0.75

0.50

0.25

0.00 0 0 25 50 75 25 50 75 100 100 Genomic Distance (kilobases)

Supplementary Figure 4: LD decay by superpopulation Scaling of LD with genomic distance shows moderate anti-correlation for all superpopulations. Figure 2 shows combined LD scaling and observed Hi-C contact frequency scaling by cell line.

32 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

AFR (ρ = − 0.64) AMR (ρ = − 0.53) EAS (ρ = − 0.6) 1.00

0.75

0.50 )

2 0.25 R

0.00 0 25 50 75 100 EUR (ρ = − 0.62) SAS (ρ = − 0.61) 1.00

0.75 Max Pairwise LD ( Max Pairwise

0.50

0.25

0.00 0 0 25 50 75 25 50 75 100 100 Genomic Distance (kilobases)

Supplementary Figure 5: Interaction LD decay by superpopulation Scaling of LD with genomic distance for SNPs located on statistically signiﬁcant chromatin interactions shows moderate anti-correlation for all superpopulations. Figure 2 shows combined LD scaling (not restricted to interacting chromatin) and observed Hi-C contact frequency scaling by cell line.

33 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

(a) Chromatin interactions from blood cell types (b) Chromatin interactions from ENCODE cell lines

AFR AMR EAS AFR AMR EAS

1e+05 1000 1e+03 1e+01 10 0−10 0−10 0−10 0−10 0−10 0−10 > 500 > 500 > 500 > 500 > 500 > 500 10−50 10−50 10−50 10−50 10−50 10−50 50−100 50−100 50−100 50−100 50−100 50−100 100−500 100−500 100−500 100−500 100−500 100−500

log(Count) EUR SAS log(Count) EUR SAS

1e+05 1000 1e+03 1e+01 10 0−10 0−10 0−10 0−10 > 500 > 500 > 500 > 500 10−50 10−50 10−50 10−50 50−100 50−100 50−100 50−100 100−500 100−500 100−500 100−500 # Crossed LD Blocks # Crossed LD Blocks

AFR AMR EAS

1e+05

1e+03

1e+01

1 2 3 4 5 6 7 1 2 3 4 5 6 7 1 2 3 4 5 6 7

EUR SAS log(Count)

1e+05

1e+03

1e+01

1 2 3 4 5 6 7 1 2 3 4 5 6 7 # Crossed Contact Domains

Supplementary Figure 6: Interaction crossings Number of LD blocks crossed by statistically signiﬁcant chromatin interactions across all chromosomes in 17 primary blood cell types [12] (panel a) and 5 ENCODE cell lines [21] (panel b), grouped by superpopulation. Many statistically signiﬁcant interactions cross hundreds or thousands of LD blocks. Panel c) shows the number of contact domains crossed by SNP pairs with non-zero LD, summed over 5 ENCODE cell lines, grouped by superpopulation.

34 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

References

1. Welter, D. et al. The NHGRI GWAS Catalog, a curated resource of SNP-trait associations. Nucleic Acids Research 42, D1001–D1006 (2014). 2. Farh, K. K.-H. et al. Genetic and epigenetic fine mapping of causal autoimmune disease variants. Nature 518, 337–343 (2014). 3. Gusev, A. et al. Partitioning Heritability of Regulatory and Cell-Type-Specific Variants across 11 Com- mon Diseases. American Journal of Human Genetics 95, 535–552 (2014). 4. Kundaje, A. et al. Integrative analysis of 111 reference human epigenomes. Nature 518, 317–330 (2015). 5. Parker, S. C. J. et al. Chromatin stretch enhancer states drive cell-specific gene regulation and harbor human disease risk variants. Proceedings of the National Academy of Sciences of the United States of America 110, 17921–6 (2013). 6. Kim, K. et al. Chromatin structure-based prediction of recurrent noncoding mutations in cancer. Nature Genetics 48, 1321–1326 (2016). 7. Nishizaki, S. S. & Boyle, A. P. Mining the Unknown: Assigning Function to Noncoding Single Nucleotide Polymorphisms. Trends in Genetics 33, 34–45 (2017). 8. Liang, S. et al. iRegNet3D: three-dimensional integrated regulatory network for the genomic analysis of coding and non-coding disease mutations. Genome Biology 18, 10 (2017). 9. Lu, Y., Quan, C., Chen, H., Bo, X. & Zhang, C. 3DSNP: a database for linking human noncoding SNPs to their three-dimensional interacting genes. Nucleic Acids Research 45, D643–D649 (2017). 10. Claussnitzer, M. et al. FTO Obesity Variant Circuitry and Adipocyte Browning in Humans. New England Journal of Medicine 373, 895–907 (2015). 11. Kirsten, H. et al. Dissecting the genetics of the human transcriptome identifies novel trait-related trans - eQTLs and corroborates the regulatory relevance of non-protein coding loci. Human Molecular Genetics 24, 4746–4763 (2015). 12. Javierre, B.-M. et al. Lineage-Specific Genome Architecture Links Enhancers and Non-coding Disease Variants to Target Gene Promoters. Cell 167, 1369–1384.e19 (2016). 13. Won, H. et al. Chromosome conformation elucidates regulatory relationships in developing human brain. Nature 538, 523–527 (2016). 14. Corradin, O. et al. Modeling disease risk through analysis of physical interactions between genetic variants within chromatin regulatory circuitry. Nature Genetics 48, 1313–1320 (2016). 15. Kim, M. J. et al. SUGP1 is a novel regulator of cholesterol metabolism. Human Molecular Genetics 25, 3106–3116 (2016). 16. Mitchel, K. et al. RP1-13D10.2 Is a Novel Modulator of Statin-Induced Changes in Cholesterol. Circu- lation: Cardiovascular Genetics 9, 223–30 (2016). 17. Mumbach, M. R. et al. Enhancer connectome in primary human cells identifies target genes of disease- associated DNA elements. Nature Genetics 49, 1602–1612 (2017). 18. He, H. et al. Multiple functional variants in long-range enhancer elements contribute to the risk of SNP rs965513 in thyroid cancer. Proceedings of the National Academy of Sciences of the United States of America 112, 6128–33 (2015). 19. Sanjana, N. E. et al. High-resolution interrogation of functional elements in the noncoding genome. Science 353, 1545–1549 (2016). 20. Dixon, J. R. et al. Topological domains in mammalian genomes identified by analysis of chromatin interactions. Nature 485, 376–80 (2012). 21. Rao, S. S. et al. A 3D Map of the Human Genome at Kilobase Resolution Reveals Principles of Chro- matin Looping. Cell 159, 1665–80 (2014). 22. Baudat, F. et al. PRDM9 Is a Major Determinant of Meiotic Recombination Hotspots in Humans and Mice. Science 327, 836–840 (2010).

35 bioRxiv preprint doi: https://doi.org/10.1101/272245; this version posted February 26, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

23. Ong, C.-T. & Corces, V. G. CTCF: An architectural protein bridging genome topology and function. Nature Reviews Genetics 15, 234–46 (2014). 24. Auton, A. et al. A global reference for human genetic variation. Nature 526, 68–74 (2015). 25. Fairfax, B. P. et al. Genetics of gene expression in primary immune cells identifies cell typespecific master regulators and roles of HLA alleles. Nature Genetics 44, 502–510 (2012). 26. Durand, N. C. et al. Juicer Provides a One-Click System for Analyzing Loop-Resolution Hi-C Experi- ments. Cell Systems 3, 95–98 (2016). 27. Cairns, J. et al. CHiCAGO: robust detection of DNA looping interactions in Capture Hi-C data. Genome Biology 17, 127 (2016). 28. Gabriel, S. B. et al. The Structure of Haplotype Blocks in the Human Genome. Science 296, 2225–9 (2002). 29. Aguet, F. et al. Genetic effects on gene expression across human tissues. Nature 550, 204–213 (2017). 30. MacArthur, J. et al. The new NHGRI-EBI Catalog of published genome-wide association studies (GWAS Catalog). Nucleic Acids Research 45, D896–D901 (2017). 31. Krefting, J., Andrade-Navarro, M. A. & Ibn-Salem, J. Evolutionary stability of topologically associating domains is associated with conserved gene regulation. bioRxiv (2017). 32. Lazar, N. H. et al. The genomic false shuffle: epigenetic maintenance of topological domains in the rearranged gibbon genome. bioRxiv (2017). 33. Quinlan, A. R. & Hall, I. M. BEDTools: A flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841–842 (2010). 34. Chang, C. C. et al. Second-generation PLINK: rising to the challenge of larger and richer datasets. GigaScience 4, 7 (2015). 35. McKinney, W. Python for Data Analysis (O’Reilly, 2012). 36. Hunter, J. D. Matplotlib: A 2D graphics environment. Computing In Science & Engineering 9, 90–95 (2007). 37. Wickham, H. ggplot2: Elegant Graphics for Data Analysis (Springer-Verlag New York, 2009). 38. Tange, O. GNU Parallel - The Command-Line Power Tool. ;login: The USENIX Magazine 36, 42–47 (2011).